Automating Vector Autoregression on Electronic Patient Diary Data.

Area of Science:

• Statistics
• Computational Science

Background:

• Vector autoregression (VAR) model selection is typically a manual, iterative process demanding significant statistical expertise and time.
• Existing automated solutions are limited and still require expert input.

Purpose of the Study:

• To introduce Autovar, a novel application designed to automate the selection of vector autoregression models for time series data.
• To demonstrate the feasibility of large-scale VAR model selection using an exhaustive approach.

Main Methods:

• Development and implementation of the Autovar application.
• Comparison of Autovar's performance against manual expert selection.
• Benchmarking Autovar against the leading commercial VAR model selection software.

Main Results:

• Autovar automates VAR model selection, mimicking expert workflows but enhanced by computational power.
• An exhaustive model selection approach is demonstrated to be effective and reliable.
• Autovar shows that large-scale vector autoregression model selection is feasible and efficient.

Conclusions:

• Autovar offers an automated, efficient, and reliable method for vector autoregression model selection.
• The successful implementation of Autovar paves the way for more accessible adaptive and personalized treatments in healthcare.
• This work advances the automation of statistical modeling for diverse applications, including medical research.